Has the Modern Bullpen Destroyed Late-Inning Comebacks?

During the World Series, I submitted an article showing that the team leading Los Angeles Dodger games after six innings wound up winning the game 94% of the time, the highest proportion in baseball. I suggested that maybe that’s why Dodger fans leave home games early; it’s a rational decision based on the unlikelihood of a late-inning rally. (Note: I was kidding. Back off.)

But between the lack of comebacks for the Dodgers and the postseason legend of Kansas City’s HDH bullpen, I wondered: Are we seeing fewer comebacks in the late innings now? The last time the Royals made the postseason, in 1985, the average American League starter lasted 6.17 innings, completing 15.9% of starts. The average American League game had 1.65 relievers pitching an average of an inning and two thirds each. This year, American League starters averaged 5.93 innings, completing just 2.5% of starts. The average game had just under three relievers throwing an average of 1.02 innings each. The reason that starters don’t go as long isn’t my point here, and has been discussed endlessly in any case. But looking at the late innings, today’s batters are less likely to be facing a starter frequently (52% of AL starters faced the opposing lineup a fourth time in 1985, compared to 28% in 2014), reducing the offensive boost from the times-through-the-order penalty. Instead, modern batters face a succession of relievers, who often have a platoon advantage, often throwing absolute gas. The Royals, as I noted, won 94% of the games that they led going into the seventh inning (65-4), and that wasn’t even the best record in baseball, as the Padres were an absurd 60-1. Are those results typical? Has the modern bullpen quashed the late-innings rally?

To test this, I checked the percentage of games in each year in which the team leading after six innings was the final victor. (These data are available at Baseball-Reference.com, using the Scoring and Leads Summary.) I went back 50 years, recording the data for every season from 1965 to 2014. In doing so, I got the 1968 Year of the Pitcher; the 1973 implementation of the DH; the expansions of 1969, 1977, 1993, and 1998; the Steroid Era; and the recent scoring lull. This chart summarizes the results:

The most important caveat here is this: In Darrell Huff’s 1954 classic, How to Lie with Statistics, he devotes a chapter to “The Gee-Whiz Graph,” in which he explains how an argument can be made or refuted visually by playing with the y-axis of a graph. This is a bit of a gee-whiz graph, in that the range of values, from 84.1% of leads maintained in 1970 to 88.1% in 2012, is only 4%. That’s not a lot. My choice of a y-axis varying from 84.0 to 88.5 magnifies some pretty small differences. Still, the average team was ahead or behind after six innings 140 times last season, so the peak-to-trough variance is five to six games per year (140 x 4% = 5.6). That’s five or six games in which a late-inning lead doesn’t get reversed, five or six games in which there isn’t a comeback, per team per season.

The least-squares regression line for this relationship is Percentage of Games Won by Team Leading After Six Innings = 85.756% + .015%X, where X = 1 for 1965, 2 for 1966, etc. The R-squared is 0.06. In other words, there isn’t a relationship to speak of. There is barely an upward trend, and the fit to the data is poor. And that makes sense from looking at the graph. The team leading after six innings won 86.5% of games in the Year of the Pitcher, 86.6% when the Royals were last in the Series, 85.7% in the peak scoring year of 2000, and 86.1% in 2013. That’s not a lot of variance. You might think that coming back from behind is a function of the run environment–it’s harder to do when runs are scarce–but the correlation between runs per game and and holding a lead after six innings, while negative (i.e., the more runs being scored, the harder it is to hold the lead), is weak (-0.22 correlation coefficient).

So what this graph appears to be saying, with one major reservation I’ll discuss later, is that the emergence of the modern bullpen hasn’t affected the ability of teams to come back after the sixth inning. Why is that? Isn’t the purpose of the modern bullpen to lock down the last three innings of a game? Why hasn’t that happened? Here are some possible explanations.

• Wrong personnel. Maybe the relievers aren’t all that good. That seems easy to dismiss. In 2014, starters allowed a 3.82 ERA, 3.81 FIP, 102 FIP-. Relievers allowed a 3.58 ERA, 3.60 FIP, 96 FIP-. Relievers compiled better aggregate statistics. A pitcher whose job is to throw 15 pitches will have more success, on average, than one whose job is to throw 100.
• Wrong deployment. Analysts often complain that the best reliever–usually (but not always) the closer–is used in one situation only, to start the ninth inning in a save situation. The best reliever, the argument goes, should be used in the highest leverage situation, regardless of when it occurs. For example, when facing the Angels, you’d rather have your best reliever facing Calhoun, Trout, and Pujols in the seventh instead of Boesch, Freese, and Iannetta in the ninth. Managers may have the right pieces to win the game, but they don’t play them properly.
• Keeping up with the hitters. Maybe the reason teams over the past 50 years have been able to continue to come back in, on average, just under 14% of games they trail after six innings is that hitters have improved at the same time pitchers have. Batters are more selective, go deeper into counts, benefit from more extensive scouting and analysis of opposing pitchers, and get better coaching. So just as they face more and better pitchers every year, so do the pitchers face better and better-prepared hitters.

My purpose here isn’t to figure out why teams in 2013 came back after trailing after six innings just as frequently as they did in 1965, just to present that they did, despite advances in bullpen design and deployment.

Now, for that one reservation: The two years in which teams leading going into the seventh inning held their lead the most frequently were 2012 and 2014. Two points along a 50-point time series do not make a trend, so I’m not saying it’s recently become harder to come back late in a game. After all, the percentage of teams holding a lead were below the long-term average in 2011 and 2013. But I think this bears watching. Pace of game, fewer balls in play, ever-increasing strikeouts: all of these are complaints about the modern game. None of them, it seems to me, would strike at the core of what makes baseball exciting and in important ways different from other sports the way that fewer late-inning comebacks would.